Efficient search for discrepant shadows in upright but not inverted images of natural objects

Evidence from schematic stimuli (Rensink and Cavanagh, 2004 Perception 33 1339 - 1358) suggests that search for discrepant shadows is slower for upright than inverted images--hypothetically owing to the discounting of shadows. We investigated search for shadows cast by real objects. Pebbles were photographed with and without cast shadows. The visual difference between all pebbles (without shadows) was estimated with the aid of a visual difference predictor (Lovell et al, 2006 ACM Transactions on Applied Perception 3 155), to allow manipulation of stimulus heterogeneity. The orientation of the target shadows was varied between 30° and 180°. Stimuli were presented upright (light-from-above) or inverted (light-from-below). Search slopes became steeper with increased heterogeneity, but this effect was much weaker for upright images. RTs were generally faster for upright images. However, at 30° they were slower, perhaps owing to the handover of processing from a relatively coarse-scaled shadow-system to other general-purpose visual routines. The results are consistent with a spatially coarse shadow module, activated when light appears to come from above, and uninvolved in any suppressive function.